Hacker News Reader

Comment by ken | Hacker News Reader

Call me old-fashioned, but if developers really want to minimize their impact on visitors, I wonder why they shifted away from using plain old log files. In the screenshot of the dashboard, "screen size" is the only attribute I see which can't be derived from a web server request log entry.

I was using Analog to analyze Apache logs back in the 1990's -- it's older than JS.

View on HN

I think it's part of a larger trend of outsourcing logic/services to outside SaaSs. Reminds me of that image posted here on HN of a spider-web of all of the SaaS products that just one company used centered around Salesforce.

For almost every SaaS one can always say "yeah but I can do it myself by doing A, B, and C" but the fact is that time is precious, and if you outsource parts of your website/product, it means for $19 a month you're getting a product that is worked on full-time by 4 engineers.

Analytics from server logging breaks in the cases where either:

- Page navigation often happens client-side (so most dynamic SPAs)

- Content is often cached and/or served from a CDN (so most non-trivial static sites)

In reality, nowadays most page loads don't ever touch the origin server, for a wide range of good reasons. Including analytics in the client-side page sidesteps that whole problem.

CDN could give you your logs though. I think cloudflare gives you analytics for example (not very detailed in the free version, but you can get more by paying).

AFAIK Cloudfront also let’s one store access logs.

> Page navigation often happens client-side

But most will still load the content from the backend, so you can count those requests.

Because modern web apps/sites don't make tons of requests to look at in logs - they load up the SPA bundle and all the magic of navigation/etc. happens in the browser, with round trips back only to fetch occasional new resources.

Because a lot of times things are hosted on services where you don't have access to raw log files (Netlify for example).

And in the Netlify example, they'll sell you analytics based on the logs (adblocker proof) for $9/month. For one case on a site I operate, the numbers are ~2x higher than GA.

This was always the way back before GA really gained traction - a lot of the old school log monitoring analytics basically equated requests as hits (with only some slight filtering) and so the numbers were always inflated (I never normally got to 2x, normally to about 1.5x - but guess it depends on the implementation's filtering of bots, spam IPs etc... Edit: remember we were using AWStats).

Was great being able to spin the traffic figures on our game review site for PRs who didn't understand web analytics at that time.

2x is usually just unique visitors, because most people block ads and trackers (at least those who visit tech blogs).

Counting all requests will be much more than 2x the GA visitors.

maybe they should not sellout their users like this and host their own files

Logs are missing some information I occasionally need to make decisions:

- Browser and screen resolution

- Time on page

- Tracking of on-page events

Reply on news.ycombinator.com